The mixed deciduous forests of the upper Midwest, USA are approaching an ecological threshold in which early successional dominant aspen and birch trees are reaching maturity and beginning to senesce. At the University of Michigan Biological Station in northern Michigan, we are combining long-term carbon (C) cycling measurements with a large-scale experimental manipulation to forecast how net ecosystem production will change in response to ongoing succession, disturbance, and climate variation. Our goal is to elucidate biophysical mechanisms that will constrain C storage in future forests. In the spring of 2008, we began the Forest Accelerated Succession ExperimenT (FASET), in which all aspen and birch ( 35% canopy LAI) within 39 ha of an 85 yr old forest were stem girdled to accelerate mortality. The adjacent, untreated forest serves as a control. I hypothesized that, 1) aspen-birch senescence would decrease C allocation to root and microbial pools, resulting in increased root mortality and reduced root respiration, 2) that the treatment effect on soil respiration (Rs) would become more severe as time since girdling increased, and 3) the magnitude of the treatment effect would be proportional to the percent basal area of girdled aspen and birch. Sites with high percent basal area of girdled aspen and birch would yield lower Rs, while sites with low percent basal area of girdled aspen and birch and control sites would have higher Rs. We measured Rs continuously using arrays of automated soil respiration chambers in the treatment and control sites. We also quantified above and below-canopy radiation, precipitation, air and soil temperature, and soil moisture. At the ecosystem scale, we calculated gross primary production from measurements of net CO2 exchange between forest and atmosphere using eddy-covariance methods. Soil respiration was significantly reduced in the treatment relative to the control site in the first and second years after girdling, supporting my first hypothesis. However, the treatment effect occurred faster than anticipated, because root C stores were expected to maintain Rs despite girdling in the first treatment year. I modeled the automated chamber Rs data using a piece-wise linear regression and found that the magnitude of the treatment effect was significantly greater in 2009 compared to 2008, supporting my second hypothesis. Results from the basal area gradient study showed no significant difference between treatment and control site Rs. However, percent basal area of aspen and birch was a significant explanatory parameter in the model, revealing that as percent basal area of aspen and birch increased in the control site, so did Rs. The treatment site showed no relationship between Rs and basal area, suggesting the treatment is beginning to deteriorate normal forest belowground functioning. My results provide an estimate of Rs after a species specific disturbance and can be extrapolated to other such disturbances such as those caused by stem girdling insects or diseases. My Rs estimates can also be applied to calculations of net ecosystem productivity, because Rs is the main component of ecosystem respiration.